IPv4 versus IPv6 - Who connects faster? Introduction Motivation Research Question Research Contributions Methodology Metrics Selection of Websites Measurement Setup Measurement Trial Data Analysis TCP Connect Times Website Clusters Comparing by CDN Special Cases Conclusion References Appendix Comparing by Network Type IPv4 versus IPv6 - Who connects faster? IFIP Networking 2015 Conference, Toulouse Vaibhav Bajpai [email protected]Jürgen Schönwälder [email protected]Computer Networks and Distributed Systems Group, Jacobs University Bremen, Bremen, Germany May 2015 1 / 32
32
Embed
IPv4 versus IPv6 - Who connects faster? - IFIP Networking 2015 ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
IPv4 versus IPv6 - Who connects faster?IFIP Networking 2015 Conference, Toulouse
▶ Returns a list of endpoints in an order that prioritizes an IPv6-upgrade path.▶ The order is dictated by RFC 6724 [2] and /etc/gai.conf
4 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Introduction | Research Contributions
▶ Websites cluster around CDN deployments and show similar performance.▶ CDN clusters are different over IPv4 and IPv6.▶ CDN caches are largely absent over IPv6.▶ TCP connect times to IPv6 CDN clusters have improved over time.▶ IPv6 services from www.bing.com have stopped globally since 2013.▶ Google CDN blacklists some resolvers over IPv6.
To the best of our knowledge, this is the first study to compare TCP connect times of topdual-stacked websites over IPv4 and IPv6.
5 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Methodology
6 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Methodology | Metrics
▶ Uses getaddrinfo(…) to resolve service names.▶ Uses non-blocking TCP connect(…) calls.▶ DNS resolution time is not accounted.▶ Can read multiple service names as arguments.▶ Can read service names list from a file.▶ File locking capability.▶ Sets a delay between connect(…) ; avoids SYN floods.▶ Can produce both human-readable & CSV output.▶ Cross-compiled for OpenWrt; Running on SamKnows.
▶ Hurricane Electric (HE) maintains a top 100 dual-stacked service names list2.▶ HE uses top 1M service names list from Alexa Top Sites (ATS)▶ HE does not follow CNAMES.
▶ Amazon has made the ATS top 1M service names list public3.▶ Prepared a custom top 100 dual-stacked service names list.▶ Explicitly follow CNAMES.▶ Prepend a www to each service name and cross-check any AAAA response.
RESIDENTIAL AS3320 AS3320 BREMEN DEUTSCHE TELEKOM #04
RESIDENTIAL AS50989 AS1257 STOCKHOLM SITAB #11
RESIDENTIAL AS4685 AS4718 FUKUOKA ASAHI NET #12
RESIDENTIAL AS12715 AS12715 MADRID JAZZ TELECOM #13
RESIDENTIAL AS9031 AS9031 ALLEUR EDPNET #17
RESIDENTIAL AS3320 AS3320 BREMEN DEUTSCHE TELEKOM #19
RESIDENTIAL AS2518 AS2516 SHIZUOKA BIGLOBE NEC #20
RESEARCH AS513 AS513 CERN CERN #16
NREN AS680 AS680 BREMEN DFN #01
NREN AS2614 AS2614 TIMISOARA ROEDUNET #08
NREN AS2611 AS2611 LOUVAIN BELNET #15
NREN AS680 AS680 BREMEN DFN #18
LAB AS5607 AS5607 LONDON BSKYB-BROADBAND #05
LAB AS3269 AS3269 TORINO TELECOM ITALIA #06
LAB AS8903 AS8903 MADRID BT ESPANA #07
LAB AS2856 AS5400 IPSWICH BT UK #10
IXP AS18070 AS18070 NIIGATA NDAC #14
BUSINESS AS24956 AS24956 BRAUNSCHWEIG GAERTNER DATENSYSTEME #03
BUSINESS AS13030 AS13030 OLTEN INIT SEVEN #09
10 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis4
4The results are derived from year-long measurements conducted between Feb 2013 - July 2014.11 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | TCP Connect Times (Probe #04)
101
102
103
TC
P c
onnect
tim
e (
mse
cs) SamKnows Probe: #04 [04-May-2013 CET to 23-May-2014 CET]
IPv4
big
lobe.n
e.jp
ww
w.a
ol.co
mw
ww
.appsp
ot.
com
ww
w.a
tt.c
om
ww
w.b
ing.c
om
ww
w.b
logger.
com
ww
w.b
logsp
ot.
co.u
kw
ww
.blo
gsp
ot.
com
ww
w.b
logsp
ot.
de
ww
w.b
logsp
ot.
inw
ww
.blo
gsp
ot.
jpw
ww
.blo
gsp
ot.
ruw
ww
.com
cast
.net
ww
w.d
eti
k.co
mw
ww
.double
clic
k.co
mw
ww
.elm
undo.e
sw
ww
.engadget.
com
ww
w.f
ace
book.
com
ww
w.f
bcd
n.n
et
ww
w.f
lipka
rt.c
om
ww
w.f
olh
a.u
ol.co
m.b
rw
ww
.fre
e.f
rw
ww
.google
.ae
ww
w.g
oogle
.at
ww
w.g
oogle
.az
ww
w.g
oogle
.be
ww
w.g
oogle
.ca
ww
w.g
oogle
.ch
ww
w.g
oogle
.cl
ww
w.g
oogle
.cn
ww
w.g
oogle
.co.h
uw
ww
.google
.co.id
ww
w.g
oogle
.co.il
ww
w.g
oogle
.co.in
ww
w.g
oogle
.co.jp
ww
w.g
oogle
.co.k
rw
ww
.google
.co.t
hw
ww
.google
.co.u
kw
ww
.google
.co.v
ew
ww
.google
.co.z
aw
ww
.google
.com
ww
w.g
oogle
.com
.ar
ww
w.g
oogle
.com
.au
ww
w.g
oogle
.com
.bd
ww
w.g
oogle
.com
.br
ww
w.g
oogle
.com
.co
ww
w.g
oogle
.com
.eg
ww
w.g
oogle
.com
.hk
ww
w.g
oogle
.com
.mx
ww
w.g
oogle
.com
.my
ww
w.g
oogle
.com
.ng
ww
w.g
oogle
.com
.pe
ww
w.g
oogle
.com
.ph
ww
w.g
oogle
.com
.pk
ww
w.g
oogle
.com
.sa
ww
w.g
oogle
.com
.sg
ww
w.g
oogle
.com
.tr
ww
w.g
oogle
.com
.tw
ww
w.g
oogle
.com
.ua
ww
w.g
oogle
.com
.vn
ww
w.g
oogle
.cz
ww
w.g
oogle
.de
ww
w.g
oogle
.dk
ww
w.g
oogle
.dz
ww
w.g
oogle
.es
ww
w.g
oogle
.fi
ww
w.g
oogle
.fr
ww
w.g
oogle
.gr
ww
w.g
oogle
.ie
ww
w.g
oogle
.it
ww
w.g
oogle
.nl
ww
w.g
oogle
.no
ww
w.g
oogle
.pl
ww
w.g
oogle
.pt
ww
w.g
oogle
.ro
ww
w.g
oogle
.ru
ww
w.g
oogle
.se
ww
w.g
oogle
.sk
ww
w.g
oogle
use
rconte
nt.
com
ww
w.irs
.gov
ww
w.m
arc
a.c
om
ww
w.m
obile
.de
ww
w.m
ozi
lla.o
rgw
ww
.netf
lix.c
om
ww
w.n
ifty
.com
ww
w.o
range.f
rw
ww
.saku
ra.n
e.jp
ww
w.s
ezn
am
.cz
ww
w.t
-onlin
e.d
ew
ww
.terr
a.c
om
.br
ww
w.u
ol.co
m.b
rw
ww
.usp
s.co
mw
ww
.vk.
com
ww
w.w
ikim
edia
.org
ww
w.w
ikip
edia
.org
ww
w.w
ikti
onary
.org
ww
w.y
ahoo.c
om
ww
w.y
oum
7.c
om
ww
w.y
outu
be.c
om
101
102
103
TC
P c
onnect
tim
e (
mse
cs)
IPv6
Boxplot grouped by service
▶ Several websites5 show similar performance over IPv4/IPv6.▶ Some disparate websites6 appear faster over IPv4.▶ Some websites7 show substantially higher variance over IPv6.
▶ Akamai caches serve popular websites directly from within DTAG over IPv4.
▶ Akamai caches missing over IPv6 (served instead by Akamai CDN).
15 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | BGP-based Website Clusters (IPv4)
0.0 0.2 0.4 0.6 0.8 1.0
Number of Services
0.0
0.2
0.4
0.6
0.8
1.0
CDF of Services within each IPv4 BGP-based Cluster
GOOGLE (AS15169)
SEABONE (AS6762)
NTT (AS2914)
JAZZNET (AS12715)
AKAMAI (AS20940)
WIKIMEDIA (AS43821)
COMHEM (AS39651)
WIKIMEDIA (AS14907)
ECHIGO (AS55895)
AOL (AS1668)
UNIDAD (AS9052)
TINET (AS3257)
EDPNET (AS9031)
TELIANET (AS1299)
UNIVERSO (AS7162)
AKAMAI (AS16625)
FACEBOOK (AS32934)
0 1 2 3 4 5 60.0
0.2
0.4
0.6
0.8
1.0
% o
f N
um
ber
of
Sam
Know
s Pro
bes:
20
60 62 64 66 68 70
IPv4 Cluster #(↓)
GOOGLE (AS15169) 67
SEABONE (AS6762) 05
NTT (AS2914) 03
JAZZNET (AS12715) 03
AKAMAI (AS20940) 03
WIKIMEDIA (AS43821) 03
COMHEM (AS39651) 03
WIKIMEDIA (AS14907) 03
ECHIGO (AS55895) 02
AOL (AS1668) 02
UNIDAD (AS9052) 02
TINET (AS3257) 02
EDPNET (AS9031) 02
TELIANET (AS1299) 02
UNIVERSO (AS7162) 02
AKAMAI (AS16625) 02
FACEBOOK (AS32934) 02
▶ CDN caches serve popular websites over IPv4 from within service provider's network.
16 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | BGP-based Website Clusters (IPv6)
0.0 0.2 0.4 0.6 0.8 1.0
Number of Services
0.0
0.2
0.4
0.6
0.8
1.0
CDF of Services within each IPv6 BGP-based Cluster
GOOGLE (AS15169)
AKAMAI (AS20940)
WIKIMEDIA (AS43821)
CW (AS1273)
WIKIMEDIA (AS14907)
UNIDAD (AS9052)
TINET (AS3257)
ECHIGO (AS55895)
DFN (AS680)
ACONET (AS1853)
UNIVERSO (AS7162)
NTT (AS2914)
FBDC (AS10013)
FACEBOOK (AS32934)
0 1 2 3 4 5 60.0
0.2
0.4
0.6
0.8
1.0%
of
Num
ber
of
Sam
Know
s Pro
bes:
20
60 62 64 66 68 70
IPv6 Cluster #(↓)
GOOGLE (AS15169) 67
AKAMAI (AS20940) 04
WIKIMEDIA (AS43821) 03
CW (AS1273) 03
WIKIMEDIA (AS14907) 03
UNIDAD (AS9052) 02
TINET (AS3257) 02
ECHIGO (AS55895) 02
DFN (AS680) 02
ACONET (AS1853) 02
UNIVERSO (AS7162) 02
NTT (AS2914) 02
FBDC (AS10013) 02
FACEBOOK (AS32934) 02
▶ CDN caches over IPv6 are largely absent.
17 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Introduction | Summary
▶ Several websites show similar performance over IPv4/IPv6.…because they cluster around a CDN.
▶ Some websites show higher variance over IPv6…because TCP connect times over IPv6 have improved over time.
▶ Some websites appear faster over IPv4…because they are served by CDN caches over IPv4.…CDN caches over IPv6 are largely absent.
18 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | GOOGLE (AS15169)
10-1 100 101 102 103 104
TCP connect time (msecs)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv6 m
easu
rem
ents
: 201544)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv4 m
easu
rem
ents
: 219346)
GOOGLE (AS15169)
IPv6
IPv4
▶ Who connects faster? - IPv6 appears comparable to IPv4.
19 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | FACEBOOK (AS32934)
100 101 102 103 104
TCP connect time (msecs)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv6 m
easu
rem
ents
: 6257)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv4 m
easu
rem
ents
: 6508)
FACEBOOK (AS32934)
IPv6
IPv4
▶ Who connects faster? - IPv6 slower to IPv4.20 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | AKAMAI (AS20940)
100 101 102 103 104
TCP connect time (msecs)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv6 m
easu
rem
ents
: 9047)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv4 m
easu
rem
ents
: 7375)
AKAMAI-ASN1 (AS20940)
IPv6
IPv4
▶ Who connects faster? - IPv6 slower to IPv4.21 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | WIKIMEDIA (AS43821)
100 101 102 103 104
TCP connect time (msecs)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv6 m
easu
rem
ents
: 8077)
0.0
0.2
0.4
0.6
0.8
1.0
CD
F(n
um
ber
of
IPv4 m
easu
rem
ents
: 8309)
WIKIMEDIA-EU (AS43821)
IPv6
IPv4
▶ Who connects faster? - IPv6 faster to IPv4.22 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | www.bing.com
12-M
ar-2
013
CET
13-Ju
l-201
3 CET
15-Ja
n-20
14 C
ET
24-F
eb-2
014
CET
04-A
pr-2
014
CET
09-M
ay-2
014
CET
19-Ju
n-20
14 C
ET100
101
102
103
TC
P c
onnect
tim
e o
ver
IPv6
(m
secs
)
www.bing.com
100
101
102
103
104
TC
P c
onnect
tim
e o
ver
IPv4
(m
secs
)
www.bing.com
IPv6
IPv4
12-M
ar-2
013
CET
13-Ju
l-201
3 CET
15-Ja
n-20
14 C
ET
24-F
eb-2
014
CET
04-A
pr-2
014
CET
09-M
ay-2
014
CET
19-Ju
n-20
14 C
ET0
50
100
150
200
250
300
350
TC
P c
onnect
tim
e o
ver
IPv6 (
mse
cs)
www.bing.com
0
200
400
600
800
1000
1200
1400
1600
1800
TC
P c
onnect
tim
e o
ver
IPv4 (
mse
cs)
www.bing.com
IPv6
IPv4
▶ No IPv6 measurements to www.bing.com starting Sep 2013.▶ Abrupt cut-off hinting towards a network policy decision.▶ Upstream resolvers do not provide a AAAA entry for www.bing.com .
▶ A Japanese probe running no IPv6 measurements to websites served by Google CDN.▶ Google used to perform AAAA prefix whitlisting [3].▶ Since World IPv6 Launch Day (2012), Google has changed this policy.
24 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Data Analysis | Google IPv6 Blacklist
http://googleipv6.vaibhavbajpai.com
▶ The whitelist has been replaced by a blacklist.▶ Eliminates the opt-in process; increases chances of reaching Google services over IPv6.
▶ A large number of blacklisted resolvers originate from Japan.
▶ Around 58% of blacklisted resolvers in JP, CN, TW.
26 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
Conclusion
▶ Websites cluster around CDN deployments and show similar performance.▶ CDN clusters are different over IPv4 and IPv6.▶ CDN caches are largely absent over IPv6.▶ TCP connect times to IPv6 CDN clusters have improved over time.▶ IPv6 services from www.bing.com have stopped globally since 2013.▶ Google CDN blacklists some resolvers over IPv6.
The tool is publicly released:http://happy.vaibhavbajpai.com
27 / 32
IPv4 versus IPv6 -Who connects
faster?
IntroductionMotivation
Research Question
Research Contributions
MethodologyMetrics
Selection of Websites
Measurement Setup
Measurement Trial
Data AnalysisTCP Connect Times
Website Clusters
Comparing by CDN
Special Cases
Conclusion
References
AppendixComparing by Network Type
References
[1] J. Czyz, M. Allman, J. Zhang, S. Iekel-Johnson, E. Osterweil, and M. Bailey, ``Measuring ipv6 adoption,'' inProceedings of the 2014 ACM Conference on SIGCOMM, ser. SIGCOMM '14. New York, NY, USA: ACM, 2014,pp. 87--98. [Online]. Available: http://doi.acm.org/10.1145/2619239.2626295
[2] D. Thaler, R. Draves, A. Matsumoto, and T. Chown, ``Default Address Selection for Internet Protocol Version 6(IPv6),'' RFC 6724 (Proposed Standard), Internet Engineering Task Force, Sep. 2012. [Online]. Available:http://www.ietf.org/rfc/rfc6724.txt
[3] J. Livingood, ``Considerations for Transitioning Content to IPv6,'' RFC 6589 (Informational), InternetEngineering Task Force, Apr. 2012. [Online]. Available: http://www.ietf.org/rfc/rfc6589.txt